I N S T I T U T E
A General Theory of Power Concentration: Demographic
Influences on Political Organization
John Gerring, Jillian Jaeger, Matthew Maguire
Working Paper
SERIES 2016:29
THE VARIETIES OF DEMOCRACY INSTITUTE
April 2016
Varieties of Democracy (V-Dem) is a new approach to the conceptualization and measurement of democracy. It is co-hosted by the University of Gothenburg and University of Notre Dame. With a V-Dem Institute at University of Gothenburg that comprises almost ten staff members, and a project team across the world with four Principal Investigators, fifteen Project Managers, 30+ Regional Managers, 170 Country Coordinators, Research Assistants, and 2,500 Country Experts, the V-Dem project is one of the largest-ever social science research- oriented data collection programs.
Please address comments and/or queries for information to:
V-Dem Institute
Department of Political Science University of Gothenburg
Sprängkullsgatan 19, PO Box 711 SE 40530 Gothenburg
Sweden
E-mail: contact@v-dem.net
V-Dem Working Papers are available in electronic format at www.v-dem.net.
Copyright © 2016 by authors. All rights reserved.
1
A General Theory of Power Concentration:
Demographic Influences on Political Organization
∗John Gerring
Professor of Political Science Boston University
Jillian Jaeger PhD Student Boston University
Matthew Maguire PhD Student Boston University
∗ This research project was supported by Riksbankens Jubileumsfond, Grant M13-0559:1, PI: Staffan I. Lindberg, V-Dem Institute, University of Gothenburg, Sweden; by Knut and Alice Wallenberg Foundation to Wallenberg Academy Fellow Staffan I. Lindberg, Grant 2013.0166, V-Dem Institute, University of Gothenburg, Sweden; as well as by internal grants from the Vice-Chancellor’s office, the Dean of the College of Social Sciences, and the Department of Political Science at University of Gothenburg. We performed simulations and other computational tasks using resources provided by the Notre Dame Center for Research Computing (CRC) through the High Performance Computing section and the Swedish National Infrastructure for Computing (SNIC) at the National Supercomputer Centre in Sweden. We specifically acknowledge the assistance of In-Saeng Suh at CRC and Johan Raber at SNIC in facilitating our use of their respective systems.
2
Abstract
Why is the exercise of political power highly concentrated in some polities and widely dispersed
in others? We argue that one persistent causal factor is demographic. Populous polities are
characterized by less concentrated structures of authority. To explain this relationship we
invoke two mechanisms: efficiency and trust. The theory is demonstrated with a wide variety of
empirical measures and in two settings: (1) cross-country analyses including most sovereign
states and extending back to the 19
thcentury and (2) within-country analyses focused on states,
counties, and localities in the United States.
3
Introduction
Since Aristotle pondered the virtues of rule by one, few, and many the question of political concentration has played a key role in debates about good governance. Many writers believe that only when power is diffused across many levels and many institutions are conditions optimal for constraining the abuse of power, achieving stability and credible commitment, ensuring property rights, and maximizing the utility of citizens with diverse values and interests (Breton 1996; Buchanan 1995; Elazar 1987; Gordon 1999; Inman & Rubinfeld 1997; Kollman 2013; North & Weingast 1989; Oates 1972; Ostrom 1971; Tiebout 1956; Weingast 1995).
Others view power dispersion more skeptically – as an invitation to special interest politics, weak government, and collective action dilemmas (Bagehot 1867/1963; Bardhan & Mookherjee 2000; Gerring & Thacker 2008; Keefer, Narayan & Vishwanath 2006; Prud’homme 1995;
Schattschneider 1942). According to a third perspective, the success of dispersion is contingent, i.e., dependent on contextual factors and on the type of decentralization being considered, or mixed, setting in motion offsetting virtues and vices with no straightforward implications for the overall quality of governance (Bardhan 2002; Enikolopov & Zhuravskaya 2007; Oxhorn, Tulchin & Selee 2004; Treisman 2007; Tsebelis 2000).
In this study, we focus on the prior question. Why is the exercise of political power highly concentrated in some polities and widely dispersed in others? At one extreme stands North Korea, where a small cadre micro-manages the personal lives of citizens with virtually no constraints. At another extreme lie polities like India, Switzerland, and the United States – along with confederations and international organizations such as the European Union, the United Nations, and the World Trade Organization – where decisionmaking power is diffused across many independent actors. What might account for the extraordinary variation we find in power concentration throughout the world today?
While the causes of democracy have received a great deal of attention from scholars (Coppedge 2012), the sources of power concentration are less often studied. Extant work on the subject focuses mostly on the vertical dimension, i.e., the balance of power between higher- and lower-level governments. Analyses are generally centered on democracies, the OECD, and the postwar era (e.g., Arzaghi & Henderson 2005; Garrett & Rodden 2003; Gibler 2010;
Hooghe & Marks 2013; Hooghe, Marks & Schakel 2010; Letelier 2005; Wibbels 2005). Many studies are limited to a small set of countries, especially those that are highly decentralized or have recently centralized or decentralized (e.g., Benz & Broschek 2013; Dickovick 2011; Eaton 2004; Eaton & Dickovick 2004; Falleti 2005, 2010; Garman, Haggard & Willis 2001; Grossman
& Lewis 2014; Manor 1999; Montero & Samuels 2004; O’Neill 2005; Wibbels 2005; Willis,
4 Garman & Haggard 1999; Ziblatt 2006).
As a complement to these focused studies we intend to broaden the theoretical and empirical purview. Our theory encompasses both the vertical and horizontal dimensions of power. It applies to all manner of polities – supra-national, national, and subnational – so long as each unit enjoys some degree of policymaking autonomy and so long as the community of individuals within that polity is beyond the number that could conveniently congregate in one location.
1The purview of the analysis extends throughout the modern era, inaugurated by the English, American, and French Revolutions.
With this set of scope-conditions, we argue that the degree of power concentration in a polity is affected by the number of people residing within that polity. The larger the population, the more likely that power will be diffused. Undergirding this relationship are two causal mechanisms: a search for greater efficiency and a quest to resolve problems of trust. Each of these mechanisms, we claim, exerts pressure on leaders and citizens of a large polity to diffuse power.
To test the relationship between size and power concentration we compile a cross- country dataset that incorporates most sovereign countries and a panel format extending as far back as the data will allow (in some cases to the early nineteenth century). We also provide a more focused examination of a single country – the United States – which offers a range of levels and types of government and a rich array of data with which to measure the degree of power concentration across subnational units.
The paper begins with a presentation of the argument. We then turn to crossnational empirical work on the subject, followed by a set of empirical tests. In the third section, we briefly review studies focused on the United States and then conduct our own tests. A speculative conclusion explores possible extensions of the argument.
I. Theory
A theoretical maximum of power concentration is achieved when a single individual or ruling group makes all important policy decisions in a polity. A theoretical minimum is harder to define.
Taking the existence of political elites as a given, we shall say that this ideal is achieved in a setting where power is widely dispersed, where numerous actors hold effective vetoes, and where rulers are compelled to abide by these limitations (e.g., by constitutional provisions that cannot easily be changed and are enforced by an active judiciary). Henceforth, near-synonyms
1 In micro-communities (e.g., bands, villages, neighborhoods, city-states) citizens may implement some form of direct democracy (Raaflaub, Ober & Wallace 2007) or rely on informal, non-institutionalized mechanisms to reach decisions (Ostrom 1990). Here, power is apt to be highly diffuse for the simple reason that face-to-face networks are convenient and there is little need to concentrate authority in the hands of a single leadership group.
5 such as centralization, concentration, and consolidation, and antonyms such as decentralization, de-concentration, diffusion, dispersion, and fragmentation, will be used interchangeably. All are understood to refer to the overall concentration/dispersion of power in a polity.
So defined, our topic intersects with, and to some extent subsumes, adjoining topics such as constitutional federalism (Filippov, Ordeshook & Shvetsova 2004; Riker 1964; Watts 1998), regionalism and multilevel governance (Hooghe, Marks & Schakel 2010), local governance (Bardhan & Mookherjee 2006), fiscal federalism (Boadway & Shah 2009; Gadenne
& Singhal 2014; Oates 1972; Rodden 2005), public administration (Dubois & Fattore 2009), separate powers (Vile 1967/1998), veto points (Tsebelis 2002), devolution (O’Neill 2000), delegation (Mookherjee 2006), direct and indirect rule (Gerring et al. 2011), consensus/majoritarian institutions (Lijphart 1999), and the size of states (Alesina & Spolaore 1997, 2003; Alesina, Perotti & Spolaore 1995; Bolton & Roland 1997; Colomer 2007; Lake &
O’Mahony 2004; Wittman 1991, 2000). Each of these topics is in some respects unique, justifying the focused approach taken by extant work. Yet, there is also considerable semantic and empirical overlap across these concepts. Indeed, they are difficult to disentangle.
Consider the distinction between vertical concentration (e.g., between higher and lower levels of government) and horizontal concentration (i.e., between political bodies at a single level of government). While useful for many purposes this distinction is not entirely clear-cut, as the two dimensions of power tend to intermingle (Hueglin & Fenna 2006). Let us consider a few examples. First, constitutional federalism is almost invariably combined with a second legislative chamber whose goal is to represent subnational polities (aka states, territories, or provinces), and whose representation is usually asymmetric with the lower house, generating a consequential division between the two chambers. Second, separate powers at the national level – between the executive and legislature – probably reinforce federalism, while federalism reinforces separate powers (Cameron & Falleti 2005: 257; Eaton 2004: 20-22). Third, both separate powers and federalism enhance the development of judicial review (Whittington 2009), and judicial review may help to entrench separate powers and federalism (Johnston 1969). More generally, it may be argued that divisions between branches at the national level enhance the power of subnational political forces, while powerful subnational forces enhance divisions at the national level. Vertical and horizontal fragmentation is mutually constitutive. From this perspective, and because of our quest for a unified theory, it seems appropriate to treat vertical and horizontal dimensions as manifestations of a single latent concept.
Although we strive for an encompassing definition of power concentration we do not
intend to envelope the adjacent concept of regime-type. Democracy, we shall assume, may be
6 achieved in highly concentrated polities (e.g., the United Kingdom, until quite recently) or highly de-concentrated systems (e.g., the United States). Likewise, undemocratic polities may be concentrated (e.g., the French and Spanish empires and contemporary North Korea) or de- concentrated (e.g., the British and Ottoman Empires and contemporary China). Of course, we recognize that the nature of a regime affects the degree to which power can be concentrated or de-concentrated at any given point in time. There are also important interactions between institutions that structure regimes and institutions that structure power concentration, complicating our ability to test arguments focused on the latter. These difficulties notwithstanding, regime-type serves as a background condition – not a constitutive element – of power concentration. We assume the causes of regime-type are not identical to the causes of concentration.
Causes
Extant research on our topic is focused mostly on the vertical dimension of power – especially constitutional federalism and fiscal federalism
2– and on proximal causes of centralization, e.g., the dynamics of party competition, the interplay between national and subnational elites, and economic crisis (e.g., Benz & Broschek 2013; Crémer & Palfrey 1999; Dickovick 2011; Eaton &
Dickovick 2004; Falleti 2005, 2010; Grossman & Lewis 2014; Manor 1999; Montero & Samuels 2004; O’Neill 2005; Strøm & Graham 2014; Wibbels 2006).
While proximal factors are obviously important, we should not lose sight of less visible distal factors that may be operating beneath the surface. This includes technology (communications, transport, military), geography, economic development, urbanization, inequality, globalization, external threats (including outright war), colonial heritage, ethnocultural diversity, and regime-type (Arzaghi & Henderson 2005; Dickovick 2011; Eaton &
Dickovick 2004; Garrett & Rodden 2003; Gibler 2010; Hooghe & Marks 2013; Hooghe, Marks
& Schakel 2010; Letelier 2005; Manor 1999; Strøm & Graham 2014; Veliz 1980; Wibbels 2005).
We suspect that most of these factors are limited in purview: they may affect some aspects of power concentration but not others. Some evidence will be offered for this conjecture, though it is not our goal to comprehensively assess all possible influences on this outcome.
32 A good deal of the literature on vertical concentration is concerned with what has been labeled the assignment problem, i.e., the appropriate allocation of specific duties and responsibilities across levels of government. Our concern is not with who does what (specifically) but rather with the overall concentration or dispersion of power in a polity.
3 We assume that ideational factors are pervasive. De-concentrated political institutions are more likely to be instigated and reproduced when actors believe they provide better governance, and the popularity of decentralization has waxed and waned over the past century, influencing the spread – and now, perhaps, the retraction – of decentralization initiatives (Manor 1999). However, to say that actors are responsive to ideas offers
7 Our goal is to home in on one distal cause that, we conjecture, is fairly universal, namely, the size of a polity, understood as the number of permanent residents within its boundaries.
4This provides the foundation for a general theory of power concentration. Before sketching that theory it is important to appreciate the changing historical role of size as a limiting condition of politics.
In the pre-modern era, population and power concentration were, by most accounts, strongly correlated (Downing 1992; Ember 1963; Ember, Ember & Russett 1997; Feinman 2011; Gledhill, Bender & Larsen 1988; McIntosh 1999; Stevenson 1968; Vengroff 1976).
5Large states and empires were highly concentrated (taking into account technological constraints operating at the time), and smaller bands and tribes less so. The reasons for this may be briefly reviewed, building on the literature cited above. First, the establishment of concentrated authority was critical to the rise of civilization, including population growth. Second, a large population fueled the concentration of the state apparatus by providing opportunities for revenue extraction, including forced labor. Third, populations were generally incorporated and controlled by military force, so a synergy existed between control and size – with strong states gaining population and weak states losing population. For all these reasons, size and power concentration were inseparable through most of human history, though it would be complicated to sort out cause and effect. (Presumably, they influenced each other in a reciprocal fashion, as suggested by this short explanatory sketch.)
In recent times, due to developments in technology, ideology, and political organization, relationships between state and society have changed in fundamental ways. First, the demographic transition plays out in ways that are difficult for states to control. It is no longer the case that centralized governance paves the way for population growth; if anything, it is the reverse.
6Second, state boundaries, once fluid, are now fairly rigid. Once a state gains international recognition its borders tend to remain fairly static through time. Inter-state conflict, though persistent, does not usually result in a dramatic realignment of borders (Zacher 2001). This means that the longstanding association between militarily strong – and presumably highly concentrated – states and expanding territory no longer exists, or barely exists. Even
an explanation that is nearly circular. A better explanation would reveal the prior causes of those ideas. Our hypothesis focuses on the size of a polity.
4 An early formulation of this theory, focused on territory rather than population, can be found in Dahl & Tufte (1973: 40).
5 Osafo-Kwaako & Robinson (2013) reaffirm this classical view, though they argue that Africa does not fit the mold. Note that the classical view is often focused on population density, whereas our focus is on total population.
However, most of the arguments about density also apply to population.
6 While aggressive population policies – either pro-natalist or anti-natalist – are fairly common, many demographers view such efforts as having only a marginal influence on the long-run growth of populations (Childs et al. 2005; van de Kaa 2006).
8 weak states survive (Diehl & Goertz 1992; Jackson & Rosberg 1982). Third, border changes, when they do occur, generally follow a logic of fission or fusion among peoples rather than the naked expropriation of territory by states. When borders realign it is usually with the approval, and often the connivance, of populations on the ground. People choose states, in addition to states choosing people.
7These historic transformations in the relationship of states and populations suggest that demographic factors may now serve as a largely exogenous force in political development, and that the direction of that influence may be very different in the modern era.
8We argue that a polity’s institutions are affected by the population that it has, or aspires to have (if the polity is not yet formed or is at a re-founding juncture). The larger the polity, the more fragmented its institutional design.
By way of entrée, let us consider an example of nested polity types: (1) United Nations (global), (2) European Union (supra-national), (3) France (national), (4) Midi-Pyrenees (regional), (5) Haute-Garonne (departmental), (6) Toulouse Métropole (metropolitan), and (7) Toulouse (commune). Because these governing units are nested within each other we can compare political organization across levels, holding constant some of the historical and cultural factors that might be expected to influence constitutional choices. A glance at our exemplars suggests that power at top levels is fairly diffuse, with numerous limitations on the exercise of executive power and most consequential decisions reserved for lower levels. Power at bottom levels is more concentrated, with fewer horizontal or vertical constraints.
9Of course, polities at lower levels face constraints from above – France is constrained by the EU just as Toulouse (commune) is constrained by Toulouse Metropole. However, our theory pertains to the internal organization of polities, not external constraints. One must also appreciate that comparisons across levels of government are fraught with complications since each level carries a different constitutional mandate, and some of the institutional features of interest to us may derive from those differing mandates. That is why most of the analyses to
7 At one point or another in history, migration accounts for the original inhabitants of all areas of the world except East Africa (the generally acknowledged point of origin of the human species). Our concern is with migration occurring over the past two centuries and its possible association with the outcomes of interest. A recent study of this subject finds that states exercised little control over their citizens’ comings-and-goings until the contemporary era, at which point democracies began to limit in-migration while autocracies limited out-migration. The net effect is a shift of citizens from democracies to autocracies (Breunig, Cao & Luedtke 2012). It is unclear whether the numbers involved are sufficient to impact the causal analyses presented in the following section. To the extent that it might, the impact of migration seems to run counter to our hypothesis and thus makes any positive effect of population on power diffusion/constraints less likely.
8 This fits with a growing body of work that regards demography as a moving force of history (Goldstone et al.
2011).
9 Fodder for this gross generalization may be found in Cole (2010: 307), Loughlin (2007), Loughlin et al. (2001).
[To scour]
9 follow in the empirical section of this study focus on comparisons across the same type of polity.
Nonetheless, it is consistent with our theory that variations across different levels of government usually correspond to variations in power concentration. The largest polities tend to be the least centralized; the smallest polities are generally the most centralized.
A schematic illustration of the theorized relationship between size and concentration is provided in Figure 1. Here, we visualize three exemplars, where the size of each shape indicates the size of its population. The large polity features many independent power centers – three at the top tier, thirteen at the second tier, and twenty-three at the lowest tier. The medium-sized polity features just two tiers, with just two independent power centers at the top tier and twelve at the bottom tier. The small polity has no independent power centers and is thus highly concentrated. The modular quality of Figure 1 corresponds neatly to our theory – though it must be stressed that this is a highly stylized representation.
Figure 1: Schematic Rendering of Polity Size and Concentration
Large Polity: De-Concentrated
Medium-sized Polity: Semi-Concentrated
Small Polity: Concentrated
To explain the apparent connection between demography and institutions we identify two likely causal mechanisms – efficiency and trust. In sketching out these mechanisms we consider the dynamic that obtains when the population of a polity grows while other conditions remain the same. In this setting we surmise that leaders face pressure to place limits on the
Subdivisions within a polity represent independent power centers, organized vertically and horizontally.
10 exercise of power, either by fragmenting power at the center (horizontal de-concentration) and/or devolving power from center to periphery (vertical de-concentration). Sometimes, these institutional reforms are undertaken in an explicit and intentional fashion, e.g., by statute or constitutional reform. At other times they occur unobtrusively, perhaps even without notice.
Efficiency
The size of a polity affects the relative efficiency of different organizational forms (Hooghe &
Marks 2013). Because of its larger tax base, as well as presumed economies of scale, a large polity can afford to develop highly de-concentrated political and administrative structures staffed with well-trained personnel. More importantly, the complexity of tasks facing a large polity may require a more diffuse structure of decisionmaking and implementation. Note that larger populations are generally more heterogeneous. Heterogeneity is manifested in the cultural sphere, i.e., in core values, ways of life, languages, religions, ethnicities, and so forth. It is also manifested in the economic sphere, as larger populations are likely to generate more diverse economies and hence a greater variety of sectors, occupations, and social strata. We shall assume that heterogeneity – cultural and/or economic – enhances the challenges of governance.
Informational inputs are a lot more complicated and one-size-fits all solutions are unlikely to be very satisfactory. From this perspective, de-concentration is not simply a luxury afforded to a larger polity but also a functional necessity.
Work across many disciplines suggests that the size of a unit enhances its complexity, and complexity stimulates greater specialization (aka division of labor). Research has identified such a relationship in sociobiology, where it is found at the level of individual organisms as well as animal societies (Bonner 1988, 2004; Jeanson et al. 2007), in human economies and societies (Carneiro 1967; Durkheim 1893/1964; Smith 1776), in firms (Chandler 1962; Chaney & Ossa 2013; Penrose 1959), and in professional associations (Campbell & Akers 1970). It seems plausible that the same law might hold true for polities, where the size of a community enhances the complexity of governance tasks. Leaders must gather information on citizens, synthesize that information into the decisionmaking process, and then implement appropriate policies. With a larger population there are more people to monitor, to comprehend, to rule, and a greater number of demands to process (even if leaders are concerned only with stability and not with representing the interests of their constituents).
Of course, greater specialization, by itself, does not mandate de-concentration of power.
(Consider insect colonies, primate colonies, and many pre-modern polities.) However, in the
modern era, and in human societies, there is good reason to suppose that specialization and de-
11 concentration go hand in hand. Research on firms suggests that this is the case (Bloom, Sadun
& Van Reenen 2012; Miller 1992: ch 4), and the same dynamics may also apply on a broader scale to polities. Miller (1992: 79) points out that “even the most autocratic executives find that they must depend on subordinates with the knowledge of ‘particular circumstances of time and place.’ This information monopoly…makes necessary some degree of delegation or sharing of decision-making authority.” Arguably, there are few benefits to specialization unless accompanied by a degree of policymaking autonomy. Vertical and horizontal deconcentration ensures that administrators possess the local knowledge they need to make the right decisions and can make adjustments to those decisions as needed (Hooghe & Marks 2013).
It follows that as a population grows leaders may be obliged to institute a division of labor among administrators and/or elected officials, with different bodies having jurisdiction over different policies and/or different peoples. Horizontal de-concentration involves the development of different institutions to serve different functions (e.g., executive, legislative, and judicial) or constituencies – e.g., native courts and colonial courts, as in the classic modality of indirect rule (Lange 2009; Morris 1972), laws that apply differentially to citizens of diverse faiths (Waldron 2002), and so forth. Vertical de-concentration is practicable wherever preferences, values, and identities are geographically aligned. Here, Pareto optimal outcomes may be attained by devolving power to subnational bodies (Alesina et al. 1995: 754; Besley & Coate 2003;
Bolton & Roland 1997: 1057-58; Breuss 2004: 40; Hooghe & Marks 2013: 181; Musgrave 1959;
Oates 1972; Rubinchik-Pessach 2005). Note that vertical de-concentration is efficient only if externalities across regions are limited. Here, too, size also makes a contribution, as a larger subnational grouping is more likely to internalize costs and benefits (Hooghe & Marks 2013:
181; Oates 2005, 357). It is therefore easier to decentralize power without introducing negative externalities in a large polity than in a small polity.
Trust
Efficiency is assessed according to material metrics of wellbeing, e.g., income, health, education, infrastructure, and the like. Trust engages a different dimension and calls forth somewhat different demands on government. Note that the problem of trust cannot be solved simply by adjusting policy outputs, and often runs contrary to considerations of efficiency.
Trust engages questions of security and respect, i.e., non-tangible goods. Trust is also forward looking; it concerns actions that may (or may not) be taken in the future. When a group distrusts government it means that they are anxious about what that government might do next.
Concentrated bodies, by their very nature, are incapable of achieving credible commitment to
12 policies that they might pursue in the future (North & Weingast 1989). As such, the problem of trust augurs for institutional solutions, i.e., constraints on the center that cannot be easily overcome, regardless of who happens to control the executive.
Research suggests that feelings toward government are affected by the size of government. “Big government” (understood here as a unit that rules over large masses of people) is distrusted while “small government” (a unit that rules over a small, and presumably local, group of people) is trustworthy. Consequently, local institutions are generally more trusted than regional institutions, regional institutions are more trusted than national institutions, and national institutions are more trusted than supra-national institutions (Berezina
& Diez-Medrano 2008; Nielsen 1981). Likewise, trust in large states is lower than in small states (Matsubayashi 2007; Turner 2011), and across localities with identical functions larger localities garner lower trust than smaller localities (Denters 2002). Size is also inversely correlated with political participation, efficacy, and satisfaction with democracy (Hansen 2014; Lassen &
Serritzlew 2011; Oliver 2000; Remmer 2010; Weldon 2006) – features that we presume are correlated with political trust. In sum, a body that stands closer in proximity to the citizen is apt to be trusted more than one that stands afar, all other things being equal.
10A polity need not be responsive or in any respect democratic in order for considerations of trust to come into play. Even when state elites are not inclined to implement citizen demands they must be cognizant of the costs of maintaining a form of political organization that is not perceived as legitimate. Illegitimacy may result in lax observance of the laws, tax avoidance, refusal of military service, and at the limit, secession – costs that even the most authoritarian ruler is obliged to reckon with.
By way of illustration, let us consider an oft-noted dynamic in the founding or reform of polities. Here, fissiparous groups may be granted a share of power, a guarantee of rights, or a constitutional settlement that assures their rights, as a condition of their agreement to join (or remain within) a larger polity (Arzaghi & Henderson 2005; Lijphart 1977; Riker 1964; Stepan
10 the lack of trust accorded to distant rulers in a centralized policy might be understood as a rational response to uncertainty; if one cannot assure that one’s policy preferences will be respected it is logical to demand a devolution of power to those who are more likely to share one’s policy preferences, or whose actions can be more effectively monitored and controlled. Or, there may be a primordial connection between polity size and legitimacy deeply rooted in human evolution, and therefore not a purely rational response to uncertainty. Arguably, the roots of group identity and corresponding suspicion of perceived outsiders is sociobiological insofar as evolutionary pressures favor cooperation among small groups that are geographically separated from one another (Hamilton 1964; Olsson, Ebert, Banaji & Phelps 2005). If so, the norm of “self-rule” is as old as group identity and follows logically from the latter. Note that local rule usually corresponds to rule by people who are perceived to be “like us,” while distant rule corresponds to rule by people who are classified as “them.” A third possibility is that the norm of self-rule, evidenced by the rise of ethno-nationalism as a global ideology, is a response to modernity and therefore neither purely rational (in the narrow sense of maximizing utility) nor sociobiological (Gellner 1983;
Hechter 2009: 292; Kedourie 1960).
13 1999). In this manner, vertical or horizontal de-concentration serves as a pre-condition for the birth or survival of a polity. Of course, changes to the structure of a polity are not always based on threats of dissolution. Many changes, especially those that do not involve constitutional features of a polity, occur in an incremental fashion and are scarcely perceptible except over long periods of time. This would include levels of revenue and expenditure and other more nuanced measures of relative power (see Sections II-III). Insofar as demography functions as a cause of anything at a macro-level it is often a subtle relationship, more apparent in spreadsheets than in newspaper headlines or history texts. Even so, the highly visible, macro- level negotiations noted above may be indicative of a pervasive political dynamic, one that affects power negotiations at every level.
II. Cross-Country Analyses
While the relationship of size to democracy is a storied topic (Anckar 2008; Dahl & Tufte 1973;
Gerring et al. 2015; Ott 2000; Veenendaal 2013), the relationship of size to political concentration is less often attended to. Five recent crossnational studies attempt to probe this relationship with non-negligible samples (Arzaghi & Henderson 2005; Garrett & Rodden 2003;
Hooghe & Marks 2013; Panizza 1999; Treisman 2006). Among these studies, country samples vary from 39 to 66 (with an average of 52), temporal coverage varies from 3 to 57 years (with an average of about 7), and four outcomes are considered – government consumption, constitutional federalism, regional authority, and fiscal decentralization – as summarized in Table 1.
Table 1: Recent Crossnational Studies
Study Countries Period Outcome Size Finding
Arzaghi & Henderson
2005 48 1975-1995 Gov consumption Population +
Territory 0
Federalism Population 0
Territory + Garrett & Rodden 2003 47 1982-1989 Fiscal
decentralization Population 0 Territory + Hooghe & Marks 2013 39 1950-2006 Regional authority Population + Territory 0
Panizza 1999 60 1975-1985 Fiscal
decentralization Population 0 Territory +
Treisman 2006 66 1993-95 Fiscal
decentralization Population 0 Territory +
MEAN or TOTAL 52 6.6 4 Population 2/6
Territory 4/6 Units of analysis: countries or country-years. Countries/Period: refers to the largest sample in which population or land area is included as part of the analysis. Finding: size is correlated with dispersed power in a cross-sectionally dominated analysis (+) or not (0).
14 All five studies test both population and territory as predictors of de-concentration.
These measures are transformed by the natural logarithm in accordance with the well- established notion that their impact on various outcomes depreciates in a sub-linear fashion.
Most analyses are cross-sectionally dominated, as the variables of interest tend to be sluggish and offer few opportunities for through-time analysis. (Where authors include both, we focus on the cross-sectional analysis.) Territory fares better as a predictor of de-concentration, achieving statistical significance and robustness (in various specification tests) in 4 out of 6 analyses, as shown in the final columns of Table 1. Population is vindicated in only 2 out of 6 analyses.
One must bear in mind that the chosen samples in these studies are relatively small and centered on the OECD, and thus un-representative of the universe of nation-states. Likewise, only four measures of de-concentration are tested, a rather meager representation of this vast – and difficult-to-operationalize – subject. Note that constitutional federalism can be defined and measured in many different ways; as a result, extant measures demonstrate little convergent validity (Blume & Voigt 2011). Fiscal decentralization, as measured by the IMF’s Government Finance Statistics, leaves aside the question of autonomy (are local revenue raisers able to set their own rates and spend money in ways they see fit?)
11Moreover, none of the studies reviewed in Table 1 address the possible causes of horizontal concentration.
Outcomes
In situations where individual measures of a concept are problematic it makes sense to draw on multiple measures. We regard this as a form of triangulation. If a relationship demonstrates persistence across many indicators, measurement error is less worrisome. Broadening the empirical field should also make the theory more falsifiable, as it has more chances to fail.
Additionally, we may be able to provide clues to the probable scope of the theory. For all these reasons, a wide-angle approach seems worthwhile.
Our analysis encompasses all facets of power concentration, including both horizontal and vertical dimensions, so long as they can be measured reliably across a large number of countries (100+), representing all regions of the world. Where multiple indicators purport to measure the same concept, we choose that indicator in which we have greatest confidence and/or which offers the most extensive coverage.
In this fashion, we arrive at a set of nineteen measures. Many are derived from the recently completed Varieties of Democracy project (“V-Dem”; Coppedge et al. 2015). Others
11 For further discussion see Gadenne & Singhal (2014), Hooghe & Marks (2013), Rodden (2004).
15 are constructed from the Comparative Constitutions Project (“CCP”; Elkins, Ginsburg &
Melton 2009), the Database of Political Institutions (“DPI”; Beck et al. 2001), the Political Constraints index (“PolCon”; Henisz 2002), Government Finance Statistics (“GFS”;
International Monetary Fund), World Development Indicators (“WDI”; World Bank various years), and Centripetalism (Gerring & Thacker 2008). Readers are referred to these sources for in-depth discussion of coding procedures and sources. One variable is constructed from original data collection by the authors, as described below.
The first set of indicators focuses primarily on vertical concentration (i.e., centralization). Federalism is understood as an institutionalized division or sharing of responsibilities between a national authority and semiautonomous regional units, usually codified in a constitution. Following Gerring & Thacker (2008: 88), polities are coded 0 if they are nonfederal (regional governments, if they exist, are granted minimal policy-making power), 1 if they are semifederal (there are elective governments at the regional level but constitutional sovereignty is reserved to the national government), or 2 if they are fully federal (elective regional governments plus constitutional recognition of subnational authority). Subnational government layers is comprised of two variables measuring whether (a) local or (b) regional governments exist, as coded by research assistants and regional experts enlisted by V-Dem.
These are added together to form a three-level index: 0=none, 1=one level, or 3=both levels.
Subnational elections measures the existence or non-existence of elections at subnational levels, as coded by country experts enlisted by the V-Dem project. Autonomous regions measures the existence/nonexistence of regions enjoying substantial autonomy from the national government, as coded by the DPI. Revenue decentralization is subnational revenue considered as share of total public revenue, based on the GFS and compiled by Enikolopov & Zhuravskaya (2007). Government consumption includes all (central) government current expenditures for purchases of goods and services, including payment of employees and most expenditures on national defense and security (but not those considered part of government capital formation), considered as a share of GDP, as compiled by the WDI.
12A second set of indicators focuses primarily on horizontal concentration at national levels. Separate powers is coded as 1 if the dominant executive (either the head of state or head of government) is directly elected, 0 otherwise, based on coding by research assistants enlisted by the V-Dem project. Divided party control measures the extent to which a single party or coalition
12 Because this variable focuses on the central government it offers an informative measure of the extent to which that government is able to extract resources and control the economy (Arzaghi & Henderson 2005; Oates 1972).
Though it does not focus explicitly on decentralization it complements our measure of revenue decentralization, offering superior coverage across countries and through time.
16 controls both the executive and legislative branches of national government, based on coding by country experts enlisted by the V-Dem project. Decentralized parties measures how decentralized the process of candidate selection for the national legislature is – specifically, the extent to which national party leaders control the process or share power with constituents and local and regional party actors, as judged by country experts enlisted by the V-Dem project.
13Judicial review attempts to judge whether any court in the judiciary has the legal authority to invalidate governmental policies (e.g. statutes, regulations, decrees, administrative actions) on the grounds that they violate a constitutional provision, as coded by country experts enlisted for the V-Dem project. Constitution length records the number of words in the constitution, as recorded by the CCP. Constitution scope judges the scope of a constitution based on the proportion of selected issues that are covered, following standard CCP categories. Constitution rigidity measures the number of actors required to approve constitutional amendments, according to the constitution, as coded by the CCP.
14Bicameralism measures the existence of two chambers in the national legislature and – if they exist – how closely matched their powers are, based on the coding of country experts enlisted by the V-Dem project. (If one chamber overshadows the other we regard this as an example of weak bicameralism.) Legislative committees measures whether the lower (or unicameral) chamber of the legislature has a functioning committee system and, if so, whether they are permanent (or special) and whether they have a strong influence on the course of policymaking, as coded by experts enlisted for the V-Dem project. Legislative fractionalization measures the probability that two randomly drawn representatives from the lower (or unicameral) chamber of the legislature will be from different parties, as measured by the PolCon dataset. Political constraints refers to “the extent to which a change in the preferences of any one actor may lead to a change in government policy” (Henisz 2002: 363), taking into account the number of independent branches of government and the preferences of each of these branches, as measured by the PolCon dataset. Checks & balances refers to “the number of veto players in a political system, adjusting for whether these veto players are independent of each other, as determined by the level of electoral competitiveness in a system, their respective party affiliations, and the electoral rules” (Beck et al. 2001), as measured by the DPI and transformed by the natural logarithm.
13 Following a long tradition of work, we regard the structure of parties as a key element of centralization (Filippov, Ordeshook & Shvetsova 2004; Riker 1972; Rodden 2005; Samuels & Shugart 2010; Wibbels 2006; Willis, Garman
& Haggard 1999), and candidate selection as a key influence on the relative centralization of parties (Gallagher &
Marsh 1988).
14 This measure of rigidity seems superior to others by virtue of offering significant variation across polities and strong country coverage. We do not regard actual constitutional changes as indicative of rigidity because of the confounding feature that more detailed constitutions – fostered by larger countries – are likely to need more frequent revision.
17 A final indicator, Capital city, encompasses both vertical and horizontal dimensions of power concentration. Here, we measure the population of the capital city in a polity as a share of that polity’s total population, transformed by the natural logarithm (Authors). This is regarded as a summary measure of concentration, following the assumption that in polities where power is concentrated both material resources and human resources will be concentrated at the center.
15Note that most of the tests that follow include a covariate measuring overall urbanization, so as not to confuse the status of the capital city with demographic conditions obtaining in the country at large.
This set of nineteen indicators of concentration offers a broad – but certainly not comprehensive – survey of the topic. Note that virtually any feature of government may be viewed as indicative of power concentration; as such, our topic is unbounded. One might, for example, examine particular policy areas to gauge which actors are actively engaged in policymaking and, accordingly, how concentrated that policy-area is. Sophisticated efforts of this nature (e.g., Hooghe & Marks 2010; Hooghe et al. 2016) are limited in coverage and thus not relevant for present purposes. Nonetheless, there is no reason in principle why finely-honed indicators could not be expanded to include a broader sample of countries.
16Summary features of the chosen outcomes are listed in Table 2. In addition to sources, we indicate coverage – number of countries, years, and observations – for each measure of concentration. We also show descriptive statistics – mean, standard deviation, and the scale of the index. Note that chosen indicators include a mix of continuous, left-censored, ordinal, and binary scales. All are re-scaled to 0-1 so that results can be easily compared.
15 Galiani & Kim (2011: 128) comment: “First, government agencies and workers are concentrated in capital cities.
Second, since governments make laws and redistribute income, capital cities may attract significant lobbying activity. To the extent that political corruption or rent-seeking behavior contributes to primacy, their impact is likely to be manifested in the growth of capital cities. Finally, capital cities may attract a disproportionate share of government resources for local infrastructure and amenities.” Note that previous work (Ades & Glaeser 1995;
Henderson 2003) focused on the population of capital cities, which seems less useful for present purposes.
16 One apparent omission deserves special mention. We do not regard electoral rules as having a predictable impact on centralization or decentralization. Note that district magnitude, probably the most important aspect of electoral law, has offsetting effects. Larger districts generally enhance the size of the party system but also enhance the role of national party elites in candidate selection. An increase in district magnitude thus decentralizes the party system while centralizing parties. Moreover, the impact of electoral laws is dependent upon contextual factors that are difficult to anticipate, and are open to change.
18 Table 2: Outcome Measures and Estimation Strategies
Measure Coverage Description Tests
Indicator Source Countries Years Obs Mean SD Scale Pooled Panel H
1. Federalism Centripetalism 160 103 6,354 0.191 0.374 Ordinal O. logit +
2. Subnational gov. layers V-Dem 171 115 16,181 0.931 0.176 Ordinal O. logit + 3. Subnational elections V-Dem 172 115 16,212 0.638 0.291 Continuous OLS RE +
4. Autonomous regions DPI 176 38 6,139 0.122 0.327 Binary Logit +
5. Revenue decentraliz. GFS 103 29 1,398 0.282 0.241 Continuous OLS RE +
6. Gov. consumption WDI 176 51 6,638 0.188 0.092 Continuous OLS RE –
7. Separate powers V-Dem 174 115 16,477 0.232 0.422 Binary Logit +
8. Divided party control V-Dem 174 115 16,106 0.549 0.266 Continuous OLS RE + 9. Decentralized parties V-Dem 174 115 16,281 0.339 0.199 Continuous OLS +
10. Judicial review V-Dem 174 115 16,403 0.543 0.293 Continuous OLS +
11. Constitution length CCP 197 225 960 0.031 0.080 Continuous OLS +
12. Constitution scope CCP 195 225 765 0.575 0.158 Continuous OLS +
13. Constitution rigidity CCP 191 225 906 0.318 0.265 Ordinal O. logit +
14. Bicameralism V-Dem 183 115 17,547 0.344 0.328 Continuous OLS +
15. Legislative committees V-Dem 172 115 12,106 0.533 0.173 Continuous OLS RE + 16. Legislative fractionaliz. PolCon 156 213 8,504 0.477 0.289 Censored Tobit RE + 17. Political constraints PolCon 165 213 14,808 0.220 0.290 Censored Tobit RE + 18. Checks & balances DPI 177 38 6,004 0.246 0.229 Continuous OLS RE +
19. Capital city Authors 186 111 19,895 0.509 0.120 Continuous OLS RE –
All variables re-scaled from 0-1. H: hypothesized relationship to polity size. O. logit: ordered logit. RE: random effects.
In the final columns of Table 2 we preview empirical tests to come. Following the distributions suggested by each scale, we adopt a variety of estimators in pooled regression tests.
Ten of the chosen outcomes are suitable for panel analysis, which is to say there is a sufficiently long time-series with significant variation over time in the outcome of interest. These outcomes will be analyzed with a random effects estimator and a lagged dependent variable. Finally, Table 2 notes the hypothesized relationship of polity size to the chosen indicator. Measures of concentration (Government consumption and Capital city) are expected to be negatively correlated with polity size, while measures of dispersion (all others) are expected to be positively correlated.
Appendix A includes a number of additional descriptive analyses pertaining to the
chosen outcomes. Table A3 shows that they are generally (but not always) correlated with one
another in the expected direction. A principal components analysis, shown in Table A4, shows
the first component explains only a small portion (about a fourth) of the variance. (For this
reason, attempting to reduce the information contained these indicators to a smaller number of
dimensions based on some version of factor analysis seems impracticable.) Evidently, these
nineteen indicators are capturing multiple dimensions of the underlying concept. This sets a
high bar to empirical confirmation. It is plausible that a single variable may be correlated by
accident with multiple highly correlated outcomes. It is less likely that a single predictor will be
correlated by accident with multiple weakly correlated outcomes.
19 Tests
In Table 3, we summarize a broad array of tests focused on population as a predictor of power concentration. To economize, we show estimates only for the key variable of theoretical interest – population (logged). Each cell thus represents a separate regression analysis – 175 in all.
Complete results from these tests are included in Appendix B and further information on definitions, coding, and sources for covariates can be found in Table A1.
Table 3: Cross-country Tests of Power Concentration
Analysis Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Panel Pooled
Population t-1 t-1 t-1 t-50 1900 1900 t-1 t-1 t-1 t-1, IV
Sample Full Full Full Full Full 2000 Imputed Electoral Full Full
1 2 3 4 5 6 7 8 9 10
Vertical
1.Federalism 1.028*** 0.362*** 1.126*** 0.986** 0.834** 0.732*** 0.277*** 1.097*** 0.643***
(0.310) (0.129) (0.307) (0.400) (0.330) (0.282) (0.068) (0.328) (0.119) 2.Subnational
gov layers 0.546** 0.513*** 0.789* 0.331 0.353 1.289* 0.335*** 0.477* 0.026**
(0.271) (0.110) (0.416) (0.282) (0.233) (0.711) (0.093) (0.256) (0.010) 3.Subnational
elections
0.037*** 0.048*** 0.038*** 0.034*** 0.031*** 0.028** 0.014*** 0.031*** 0.003*** 0.038***
(0.010) (0.010) (0.009) (0.010) (0.010) (0.014) (0.003) (0.010) (0.001) (0.012) 4.Autonomous
regions 0.555** 0.496*** 0.482* 0.718** 0.419* 0.687 0.404*** 0.543* 0.170
(0.267) (0.151) (0.255) (0.296) (0.225) (0.456) (0.113) (0.286) (0.138) 5.Revenue
decentraliz. 0.080*** 0.048*** 0.073*** 0.099*** 0.069*** 0.033*** 0.010*** 0.083*** 0.005*** 0.123***
(0.011) (0.015) (0.012) (0.014) (0.014) (0.010) (0.002) (0.012) (0.002) (0.021) 6.Government
consumpt. -0.014*** -0.014*** -0.013*** -0.005 -0.010*** -0.003 -0.009*** -0.009** -0.001** -0.014***
(0.004) (0.002) (0.003) (0.004) (0.003) (0.004) (0.002) (0.004) (0.000) (0.005) Horizontal
7.Separate powers
0.416*** 0.192*** 0.357** 0.314* 0.269** 0.373** 0.086 0.541** 0.174*
(0.139) (0.070) (0.165) (0.161) (0.114) (0.177) (0.065) (0.227) (0.091) 8.Divided
party control 0.023** 0.021*** 0.026*** 0.030*** 0.024** 0.045** 0.005* 0.029*** 0.002** 0.017 (0.009) (0.007) (0.010) (0.011) (0.009) (0.018) (0.003) (0.011) (0.001) (0.013) 9.Decentraliz
parties 0.026*** 0.021** 0.025** 0.022** 0.024*** 0.019** 0.010*** 0.030*** 0.021**
(0.008) (0.009) (0.010) (0.009) (0.008) (0.008) (0.003) (0.009) (0.009) 10.Judicial
review
0.021* 0.022** 0.018 0.006 0.020* 0.021* 0.012*** 0.027** 0.025*
(0.012) (0.011) (0.013) (0.013) (0.011) (0.012) (0.003) (0.012) (0.014) 11.Constitut.
length 0.000*** 0.000 0.000*** 0.000* -0.000 0.000** -0.000
(0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000)
12.Constitut.
scope 0.022*** 0.012* 0.020*** 0.012* 0.006*** 0.026*** 0.001
(0.006) (0.006) (0.007) (0.006) (0.002) (0.009) (0.008)
13.Constitut.
rigidity 0.216*** 0.024 0.051 0.155* 0.029 0.262*** -0.054
(0.073) (0.045) (0.085) (0.081) (0.018) (0.099) (0.082)
14.Bicameral-
ism 0.061*** 0.055*** 0.059*** 0.059*** 0.053*** 0.053*** 0.019*** 0.078*** 0.005*** 0.054***
(0.010) (0.009) (0.013) (0.016) (0.011) (0.017) (0.003) (0.011) (0.001) (0.013) 15.Legislative
committees
0.027*** 0.034*** 0.024*** 0.027*** 0.027*** 0.025*** 0.018*** 0.033*** 0.001*** 0.024***
(0.007) (0.007) (0.006) (0.007) (0.007) (0.006) (0.003) (0.008) (0.000) (0.009) 16.Legislative
fractionaliz 0.021* 0.017 0.023*** 0.015 0.022* 0.047*** 0.012*** 0.011*** 0.002* -0.002 (0.013) (0.017) (0.009) (0.012) (0.012) (0.015) (0.002) (0.001) (0.001) (0.016) 17.Political
constraints 0.037*** 0.077** 0.028*** 0.025*** 0.029* 0.065*** 0.013*** 0.019*** 0.002** -0.024 (0.002) (0.031) (0.001) (0.002) (0.017) (0.020) (0.002) (0.001) (0.001) (0.028) 18.Checks &
balances
0.022*** 0.008 0.015*** 0.032*** 0.022*** 0.035*** 0.014*** 0.017* 0.004*** 0.018*
(0.007) (0.009) (0.005) (0.008) (0.006) (0.010) (0.003) (0.009) (0.001) (0.010) Vertical/Horizontal
19.Capital City
-0.033*** -0.034*** -0.026*** -0.029*** -0.028*** -0.023*** -0.033*** -0.037*** -0.000** -0.035***
(0.003) (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) (0.003) (0.000) (0.004) Covariates
Basic ü ü ü ü ü ü ü ü ü
Additional ü
Yt-1 ü
Outcome measures of power concentration (re-scaled from 0-1) regressed against population (logged) and selected
